Monitoring variability in cash-crop yield caused by previous cultivation of a cover crop under a no-tillage system

Computers and Electronics in Agriculture 142 (2017) 607–621

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag

Original papers

Monitoring variability in cash-crop yield caused by previous cultivation of a cover crop under a no-tillage system

T

⁎

Júnior Melo Damiana, , Antônio Luis Santib, Michele Fornaric, Clovis Orlando Da Rosb, Vinícius Luiz Eschnerc a

Department of Soil Science, ESALQ, University of São Paulo, SP, Brazil Department of Agricultural and Environmental Sciences, Federal University of Santa Maria, RS, Brazil c Department of Agricultural and Animal Sciences, University of Passo Fundo, RS, Brazil b

A R T I C L E I N F O

A B S T R A C T

Keywords: NDVI Unmanned aerial vehicle Portable ground sensor Black oats Soybean

Cover crops may be a factor in the variability of soil fertility in areas under a no-tillage system (NTS); one which is not taken into account when fertilising the cash crop. The aim of this work was to evaluate the effect of the variability caused by the dry matter and nutrient cycling of the cover crop on the yield of the succeeding cash crop, by means of the NDVI, calculated by a portable ground sensor and using an unmanned aerial vehicle (UAV). The study was carried out in two areas of a Oxisol cultivated under NTS, where soil samples were taken, for chemical analysis and to evaluate the dry matter and nutrient accumulation in a winter cover crop of black oats. NDVI readings were taken at the R5 and R5.5 stages in soybean. A portable sensor was used seeking a comparison with the readings made by the UAV. It could be seen from the results, that the cover crop has an influence on the main crop under NTS. In the present case, the dry weight of the black oats as well as the accumulated nutrients of nitrogen, phosphorus and magnesium, showed the highest correlations with grain yield in the soybean, whereas for the chemical properties of the soil under evaluation, organic matter had the most influence on the grain yield. The NDVI assessed using the portable ground sensor and with the UAV, was efficient in evaluating the effects of variability in the crop of black oats and in the cycled nutrients.

1. Introduction Measuring the variability of various parameters of the soil, crops and production environment has always been the aim of studies into the different areas of knowledge in the agricultural sciences, with the use of new technologies providing the tools to manage and understand this variability (Jones et al., 2015; Damian et al., 2016). In Brazil, the no-tillage system (NTS) represented, and still represents, the main innovation in agriculture, in which the entire production system of the country was modified to meet the conservationist practices of this system. According to Cerri et al. (2009), NTS is a cropping system whereby the soil is turned only in the sowing furrow, and where some of the main interests are related to ground cover with suitable crops, or to crop residue from the previous harvest that are left on the surface. To this effect, the adoption of NTS has culminated in numerous studies that have sought to measure the effects of cover crops on rotation schemes under NTS, in which the results pointed to benefits that cover the physical (Costa et al., 2011; Kondo et al.2012), chemical (Torres et al., 2008; Bressan et al., 2013) and biological quality of the

⁎

soil (Santos et al., 2008; Cunha et al., 2012), as well as environmental benefits from the lesser use of mineral fertilisers (Boer et al., 2007; Crusciol and Soratto, 2009) and from reducing the effects of carbon in the environment (Amado et al., 2001; Carvalho et al., 2014). However, despite the innumerable benefits that cover crops provide at the level of an agricultural production system, these have become yet another factor in variability that need to be taken into account under NTS management (Teixeira et al., 2016). Fertilisation management is usually performed based on soil samples, after which the cover crop is planted and fertilisation carried out for the main crop only at the end of the cover-crop cycle. However, as already discussed, crop residue, especially from cover crops, can be a source of variability in the soil that is not taken into account in the fertilisation process. With this approach, the use of technology can be of great value in assessing the variability caused by cover crops on the main crop. Based on this, the NDVI (Normalized Difference Vegetation Index) may be one alternative, since it displays a correlation with crop yield (Boken and Shaykewich, 2002; Sultana et al., 2014; Lopresti et al., 2015; Peralta et al., 2016), besides serving as an indicator of numerous

Corresponding author. E-mail address: [email protected] (J.M. Damian).

https://doi.org/10.1016/j.compag.2017.11.006 Received 12 July 2017; Received in revised form 21 October 2017; Accepted 6 November 2017 Available online 20 November 2017 0168-1699/ © 2017 Elsevier B.V. All rights reserved.

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

average annual precipitation of between 1900 and 2200 mm. The distribution of precipitation, and the maximum and minimum daily temperatures during the black oat and soybean crop cycles in areas 1 and 2, are detailed in Fig. 2. The terrain in the region is gently undulating, with the soil in the two areas classified as a Oxisol (Typic Hapludox). The management used in the areas has included the adoption of a no-tillage system (NTS) for more than 20 years, and the use of PA tools, such as geo-referenced soil sampling and geo-referenced harvest monitoring.

other anomalies that may indicate the variability caused by cover crops. The NDVI can be employed by means of portable sensors placed on machines, from satellite images (Tarnavsky et al., 2008), and more recently by sensors carried on so-called unmanned aerial vehicles (UAV). UAVs are noteworthy as the most-recent technology to be employed in agriculture; but despite this, the technology has emerged as one of the main innovations in agriculture, whose use is rapidly on the increase, and the subject of worldwide research, where some studies already show the potential of this technology for use in agriculture, as is the case with studies into the control of invasive plants (Granados et al., 2016), disease (Calderón et al., 2014) and irrigation management (Dugo et al., 2013; Bellvert et al., 2014). The use of methodologies that seek to measure, understand and manage the variability that residue from cover crops causes in the main crop can represent a great advance in the management of a complex system such as NTS, mainly through the use of vegetation indices, as is the case of the NDVI. Based on this, the aim of the present study was to evaluate the effect of variability caused by the dry matter and nutrient cycling of a cover crop on the yield of the succeeding main crop, using the NDVI calculated by a portable ground sensor and a UAV.

2.2. Sampling plan for soil and plants The areas were geo-referenced using a Garmin® Legend GPS (Garmin International, Inc., Olathe, KS, USA). A sample grid with a cell size of 70.71 × 70.71 m (0.5 ha) was established for the two areas with the aid of the CR-Campeiro 7 Software (Giotto and Robaina, 2007), which interpolates a central coordinate for each pixel based on the grid size, and which resulted in 147 and 76 sampling points for Areas 1 and 2 respectively (Fig. 1). 2.3. Soil and plant properties

2. Material and methods Soil samples for chemical analysis were taken from the 147 (Area 1) and 76 (Area 2) geo-referenced points after a harvest of maize and before sowing the winter cover crop, on 15 April 2015 (Area 1) and 24 April 2015 (Area 2). The soil samples were collected from the 0.00–0.10 m layer with a quadricycle equipped with a threaded bit, powered by hydraulic drive. This depth was chosen for sampling as it is the recommended depth for areas under NTS, and has been historically used in the study areas. Fourteen sub-samples were taken in a 10 m radius of the central point of each sample cell, to make up a composite sample. After collection, the samples were identified and sent to the Soil Laboratory of the Federal

2.1. Characterisation of the study area The study was carried out using crops of black oats (Avena strigosa Schreb.) and soybean [Glycine max (L.) Merrill] o in two areas located in the town of Boa Vista das Missões in the State of Rio Grande do Sul (RS), Brazil, during 2015/2016 (Area 1) and 2016/2017 (Area 2), with Area 1 corresponding to 73.96 ha and Area 2 to 38.46 ha (Fig. 1). The climate in the region is subtropical humid with hot summers, type Cfa (Alvares et al., 2013), with a maximum temperature equal to or greater than 22 °C, a minimum temperature between −3 and 18 °C, and an

Fig. 1. Geographical location of the experimental areas used in the study.

608

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Fig. 2. Minimum daily temperature ( ), maximum daily temperature ( ) and daily precipitation ( ) during the crop cycle of the black oats (a) and soybean (b) in Area 1, and of the black oats (c) and soybean (d) in Area 2.

5445 cultivar with 220,000 seeds ha−1 (Area 1), and the P95R51 cultivar with 320,000 seeds ha−1 (Area 2), at a spacing of 0.5 m between rows in both areas. Before sowing, the seeds were treated with Fludioxonil + metalaxyl – M at a dose of 100 ml 100 kg−1 seeds (fungicide), and 200 mL thiamethoxan at a dose of 100 kg −1 seeds (insecticide). The soybean was harvested on 15 March 2016 (Area 1) and 20 February 2017 (Area 2), using a CASE® Axial-Flow 2399 harvester, containing the AFS® system of precision farming (Advanced Farming System). The system consisted of an impact-plate instant grain sensor installed at the end of a clean-grain elevator, together with a moisture sensor to determine the dry mass of the grain at 13% moisture. In addition, the harvest monitor associated information on displacement speed with platform width, and by means of a GPS signal, stored georeferenced information on the grain yield.

University of Rio Grande do Sul, where the soil organic matter content (SOM) was determined through carbon oxidation by a sulfochromic solution; pH in water (1:1); available P and K content, extracted by Mehlich-1 solution; exchangeable Ca, Mg and Al, by KCl 1 mol L−1; potential acidity (H + Al), estimated by the SMP index. The cation exchange capacity (CEC), base saturation (V) and aluminum saturation (m) were also calculated. The clay content at all points in the two study areas was greater than 60% (clayey texture). All the analyses were carried out following methodologies described in Tedesco et al. (1995) and recommended by the Commission for Soil Chemistry and Fertility – RS/SC (Commission…, 2016). Dry matter from the black oat crop at physiological maturity was then collected from the same geo-referenced soil sampling points. The black oats were sown, and samples taken, on 29 April and 29 October 2015 for Area 1, and on 03 March and 17 September 2016 for Area 2 respectively, for which a 0.25 m2 frame was used, with three replications per sampling point. It should be noted that at the time of sowing, and during the cycle of the black oats, no cropping treatments or even fertilisation were carried out, to avoid confusing the results, and to simulate actual field conditions under the management with cover crops. The samples were then taken to the Laboratory for Plant Tissue Analysis of the Federal University of Santa Maria, Frederico Westphalen campus, and placed in an oven at a temperature of 65 °C to constant weight, and then weighed on a balance with an accuracy of 0.01 g; the values were extrapolated to kg ha−1. The dry samples were ground in a Wiley-type mill with a 1.0 mm mesh screen. The nitrogen (N), phosphorus (P), potassium (K), calcium (Ca) and magnesium (Mg) were extracted from the dried and powdered material by digestion with concentrated H2SO4. The N was quantified by distillation (semi-micro Kjeldahl) and titration with sulphuric acid, the P by spectrophotometry, the K by flame photometry, and the Ca and Mg by atomic absorption spectrophotometry (Silva, 2009). After management of the black oats, the soybean crop was sown on 16 November 2015 (Area 1) and 10 October 2016 (Area 2) using the NS

2.4. Measurement of the NDVI The NDVI was used to check the effect of variations in the soil chemical properties, the dry matter yield and the cycled nutrients from the cover crop (black oats) on the main crop (soybean). The NDVI was evaluated by two methods, one using a remotely piloted aircraft system (UAV) (Fig. 3) and the other from ground readings with a GreenSeeker™ portable sensor (NTech Industries Inc. Ukiah, CA, USA). Flights with the UAV were only carried out between 12:00 and 14:00 on sunny days under a cloudless sky. The aircraft used was the fixed-wing, “Zangão” model, produced by Skydrones (Fig. 3a), which is operated by a central control system by telemetry (Fig. 3b). The maximum speed is close to 34 m/s, but it can reach higher speeds when diving, with the maximum speed of the aircraft on a mission being decided by the GSD (Ground Sample Distance) and camera shutter speed. This model is equipped with two LiPo batteries of 3300 mAh and 11.1 V containing 3 cells that resulting in a flight autonomy of 28 min. The necessary flight time will depend on the GSD, lateral overlap and 609

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Fig. 3. Model of the UAV used for evaluation (a), central control by telemetry (b) and the multispectral sensor (b) with its location in the aircraft (c).

flight speed. To obtain the spectral data, the MicaSence RedEdge multispectral sensor (MicaSence Industries Inc., Seattle, WA, USA) was used (Fig. 3c and d), which produces spectral images with channels in the blue (480 nm), green (560 nm), red (670 nm), red edge (720 nm) and near IR (840 nm) bands. The spectral parameters are related to the angular velocity, shutter speed, flight height, sensor resolution and focal length of the lens. During the flight, the stability required to obtain greater quality images is related to speed (< 24 m/s) and the angular displacement. However, it is worth mentioning that regardless of displacement of the image the drag cannot be greater than 1/2 pixel. In addition, the propeller of aircraft engines must be well balanced to prevent interferences in the quality of images. The aircraft was programmed to fly at an altitude of 200 m above ground at a speed of 20 m/s. This flight setup resulted in pixel sizes of 4.88 cm with drag of 2 cm pixel color (i.e., value lower than 1/2 of the pixel) which did not change the actual color of pixels. Flights were carried out for the soybean crop only, seeking to verify the effects of the variability of the cover crop directly on the main crop. As a result of some days not being suitable for flight (precipitation, the presence of clouds, extremely misty), two flights were made during the soybean cycle in each area, at the R5 stage (grain filling) and the R5.5 stage (pods with 50% and 75% granulation), on 28 January 2016 and 22 February 2016 (Area 1) and 12 December 2016 and 02 March 2017 (Area 2). Image processing, elaboration of the orthomosaic, and calculation of the NDVI were done using the Pix4Dmapper Pro Software (Pix4D SA Industries Inc., Lausanne, SW). Processing takes place by triangulation, where the software identifies the same point in various photos based on the coordinates of each photo, sensor resolution and lens characteristics. This same process is carried out for the other points used to generate the grid, with the precision being dependent mainly on the

quality of the image, the precision of the coordinate in the photos, and the number of anchor points in the photos. To reduce, filter and facilitate later analysis, the data were submitted to the CR-Campeiro 7 Software, where the same grid used in the previous evaluations (70.71 × 70.71 m) was applied with coinciding points, to extract the NDVI values. With this technique, it is also possible to increase precision between the compared values, because it combines a significant number of points for comparison of the results; it can also be used in other situations where the number of points and the size of the grid are maintained. The NDVI was calculated as per Eq. (1).

NDVI =

(NIR−R) (NIR + R)

(1)

where NIR is the near infrared band and R is the red band. At the same time as the two flights with the UAV, readings were taken with the portable sensor in both areas, seeking a comparison with the readings made by the UAV. This device uses light-emitting diodes in the red and near IR bands. The reflectance reading is calculated by an internal microprocessor, which gives the NDVI according to Eq. (1). The readings with the portable sensor were taken at the 147 (Area 1) and 76 (Area 2) points previously marked out by the sampling grid, between 0.8 and 1.0 m above and parallel to the ground, with five replications made in a 10 m radius of each central point, each replication comprising the representative readings of five linear metres. 2.5. Data analysis The data were initially submitted to exploratory analysis (descriptive statistics), with the aim of verifying the position and dispersion of the data using the Statistical Analysis System (SAS) 9.4 Software Package for Windows 8 (SAS Inc, Cary, EUA). The statistical parameters determined were: the minimum, mean, maximum and standard 610

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

deviation, and the coefficients of variation (CV%), asymmetry (Cs) and kurtosis (Ck). Based on the values for CV(%), data dispersion was classified as: low for a CV < 15%, moderate for a CV of 15–35%, and high for a CV > 35% (Wilding and Drees, 1983). The existence of a central tendency (normality) of the original data was also found using the W-test (p < 0.05), where data with a p-value < 0.05 was considered of normal distribution. Since most of the variables did not display a normal distribution, Spearman’s correlation was used to observe the similarity between the soil properties, the soybean crop, the black oats and the readings with the portable ground sensor and those with the UAV evaluated at stages R5 and R5.5 in the soybean in the two areas. This type of correlation is a nonparametric statistical technique analogous to the Pearson correlation coefficient, being most indicated in cases where the data do not follow a parametric distribution (Passari et al., 2011). To select from among the variables under analysis those that actually influence grain yield in the soybean, stepwise multiple regression (Y = a + b1x1 + b2x2 + … + bnxn) was applied. This functions by the systematic addition or removal of variables in the regression, carried out based on a statistical test of significance for each variable, in this study adopting a significance level of 5%, which includes in the final model only those variables that have a decisive influence on the dependent variable. The equations were evaluated according to the coefficient of determination, residual standard error and the DurbinWatson test. The Durbin-Watson test ranges from 0 to 4, where values closer to 2 indicate optimal values (Neter et al., 1985), i.e. the absence of autocorrelation in the data. The variables selected by this method were submitted to the remaining evaluations. The geostatistical analysis was carried out using experimental semivariograms, with adjustments made by means of theoretical models (spherical, exponential, gaussian and linear) using the GS+ (Gamma Design Software, LLC, Plainwell, EUA). Adjustment of the models was based on the best value for coefficient of determination (r2) and the lowest residual sum of squares (SQR), and was confirmed by the technique of cross-validation. By fitting a mathematical model to the data, the following parameters were defined: nugget (C0), contribution (C1), sill (C0 + C1) and range (a). After analysing the semivariograms for the data, and noting the spatial dependence between samples, the SURFER Software (Surfer 11, Golden Software, Inc. Golden, CO, USA) was used for spatialisation of the results using thematic maps. Interpolation was by simple kriging, taking into account the parameters of the semivariogram (adjusted model, nugget effect, range and sill) determined in the geostatistical analysis.

FPI = 1−

NCE =

c ⎡ 1 1− (c−1) ⎢ n ⎣

n ⎡ 1 − n−c ⎢ n ⎣

n

n

c

∑∑ k=1 i=1 c

∑∑ k=1 i=1

⎤ (uik )2⎥ ⎦

⎤ uik log a (uik )⎥ ⎦

(2)

(3)

where c = centroid values for the cluster; uik = values for each observation K and cluster i; loga = any positive integer; n = number of analysed data. The chosen configurations were a measure of Euclidean similarity: fuzziness exponent = 1.3, maximum number of iterations = 300, convergence criterion = 0.0001, minimum number of zones = 2, and maximum number of zones = 8. In order to discriminate the degree of differentiation between the management zones once defined, 10 sampling points, representative of seven replications, were selected for each zone, and Tukey’s test was applied (p ≤ 0.05). 3. Results and discussion 3.1. Exploratory data analysis From the descriptive statistical analysis of the soil chemical properties for the two areas (Table 1), it can be seen that only the P, SOM and Ca content, and values for CEC and base saturation in Area 1, and the values for pH, and for the K and SOM content for Area 2 showed a normal distribution of the data, explained by the coefficients of asymmetry and kurtosis being close to zero. It should be noted that normality of the data is not a requirement, but the presence of asymmetric distribution, with many anomalous values when using analyses that include linear models, should be observed more closely (Webster and Oliver, 2007). However, more important than normality of the data is the occurrence or not of the so-called proportional effect, where the mean and variance of the data may not be constant in the study area (Cavalcante et al., 2007). For data dispersion, it was found in Area 1 that the properties related to pH, SOM, Ca, Mg, CEC and base saturation displayed low dispersion in the study area (CV < 15%), where the values ranged from 3.97 to 11.89% (Table 1), and only the properties related to exchangeable Al and Al saturation had a dispersion classified as high (CV > 35%), with values from 78.57 to 85.00% respectively. In Area 2, the properties related to pH, K, SOM and Ca displayed a dispersion classified as low (CV < 15%), ranging from 4.97 to 14.20%, and only Al, Al + H and Al saturation showed a dispersion classified as high CV > 35%), ranging from 62.17 to 88.77%. Based on the values for pH (< 5.5), it was found that for both areas, correction for acidity was necessary in places (Commission…, 2016) (Table 1). As already pointed out by Anjos et al. (2012), areas of Latosol require greater attention regarding acidity, since this occurs naturally due to heavy weathering and also due to the adopted management practices. Although the values for pH are classified as having low dispersion, it should be noted that these values refer to their logarithmic scale, which may influence comparisons with other variables that do not have this peculiarity (Anjos et al., 2012). The values for available P (Mehlich-1 method) were classified as low to very high in Area 1, and from moderate to very high in Area 2, considering the textural class > 60% clay (Commission…, 2016), indicating a variation of this nutrient in the soil, as already noted above (Nanni et al., 2011; Dalchiavon et al., 2012; Ferraz et al., 2012) (Table 1). For available K, the values were classified as moderate to very high in Area 1 for CEC values between moderate and high, and with values from high to very high in Area 2 for a CEC classification at pH7.0 of between low and high. The high values for K when compared to P, are linked to the history of successive fertilisation using formulations with high concentrations of the nutrient.

2.6. Management zones The management zones were designed with the aim of defining uniform areas for the management of those variables that influence grain yield in the soybean, seeking management intervention in the short, medium or long term. In setting up the management zones, the Management Zone Analyst (MZA) 1.0.1 Software (Fridgen et al., 2004) was used. This software uses fuzzy c-means clustering algorithms whose purpose is to partition spatial observations into c-groups or clusters. The term “fuzzy” refers to the shared association between classes. The fuzziness performance index (FPI) and normalized classification entropy index (NCE) were used to determine the ideal number of clusters as well as their overall performance. The FPI (Eq. (2)) is a measure of the degree of the different classes of association (imprecision), with values ranging from 0 to 1 (Odeh et al., 1992). Whereas the NCE (Eq. (3)) is used to decide how many clusters are best suited for definition of the management zones (Bezdek, 1981). The ideal number of clusters occurs when the two indices are minimal (Fridgen et al., 2004). 611

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Table 1 Descriptive statistics of soil chemical properties evaluated in the 0–0.10 m layer. Variable§

Mínimum

Mean

Maximum

CV%†

SD‡

Cs¶

Ck#

W‡‡

Area 1 pHágua P (mg dm−3) K (mg dm−3) SOM (g dm−3) Al (cmolc dm−3) Ca (cmolc dm−3) Mg (cmolc dm−3) Al + H (cmolc dm−3) CEC (cmolc dm−3) V (%) m (%)

4.80 5.30 96.0 2.80 0.00 4.00 1.70 3.50 10.20 42.00 0.00

5.29 13.30 171.00 3.70 0.20 5.50 2.30 6.10 14.30 57.50 2.75

5.90 22.00 353.00 4.80 0.60 7.50 3.00 9.70 17.90 73.00 8.50

3.97 27.66 25.21 11.89 85.00 11.81 11.73 21.14 9.37 10.99 78.57

0.21 3.68 43.11 0.44 0.17 0.65 0.27 1.29 1.34 6.32 2.19

0.09 −0.09 1.21 0.02 0.08 0.28 0.51 0.41 0.15 −0.01 0.28

−0.33 −0.27 2.48 −0.55 −1.04 0.44 0.06 0.09 0.22 −0.18 −0.74

0.93* 0.99 ns 0.93* 0.98 ns 0.90* 0.98 ns 0.96* 0.96* 0.99 ns 0.99 ns 0.92*

Area 2 pHágua P (mg dm−3) K (mg dm−3) SOM (g dm−3) Al (cmolc dm−3) Ca (cmolc dm−3) Mg (cmolc dm−3) Al + H (cmolc dm−3) CEC (cmolc dm−3) V (%) m (%)

4.60 8.7 151.00 3.20 0.00 3.90 1.20 0.10 6.14 44.00 0.00

5.10 15.49 220.15 3.62 0.40 5.85 2.09 3.60 12.14 55.05 4.68

5.80 28.00 305.00 4.20 1.20 8.30 3.30 9.70 17.30 74.00 15.10

4.97 27.03 14.20 5.83 62.17 12.70 15.93 88.77 28.52 25.63 66.76

0.25 4.18 31.27 0.21 0.25 0.74 0.33 3.19 3.46 19.23 3.12

0.02 0.87 0.12 0.05 0.46 0.50 0.51 0.12 −0.00 −0.04 0.66

−0.49 0.61 0.01 −0.19 0.66 0.72 1.43 −1.76 −1.75 −1.82 0.82

0.96 ns 0.93* 0.99 ns 0.97 ns 0.94* 0.96* 0.96* 0.79* 0.84* 0.82* 0.95*

§ Where: P = phosphorus; K = potassium; SOM = soil organic matter; Al = exchangeable acidity; Ca = calcium; Mg = magnesium; H + Al = potential acidity; CEC = cation exchange capacity; V = base saturation; M = aluminum saturation. † Coefficient of variation. ‡ Standard deviation. ¶ Coefficient of asymmetry. # Coefficient of kurtosis. ‡‡ Shapiro-Wilk test for normal distribution. * Significant at levels of p < 0.05. ns Not significant. When significant, indicates that the hypothesis of a normal distribution is rejected.

even in the case of an area with a high use of technology in managing the variability of factors of the production system, as is the case of this study (Table 2). The DM yield of the black oat crop ranged from 2227.20 to 7722.93 kg ha−1 and from 9821.73 to 2029.47 kg ha−1, for Areas 1 and 2 respectively (Table 2). According to Da Ros and Aita (1996), Gonçalves et al. (2000) and Aita et al. (2001), in studies seeking to quantify yield in the same crop under NTS, they found average values of 3784, 3900 and 4417 kg ha−1 respectively, however as seen in this study where the distribution of DM was more-thoroughly analysed (sampling grid), these values can reach a range of from 2000 to 9000 kg ha−1, which definitely indicates DM as a factor for variability in areas under NTS. Based on the mean values for nutrients accumulated in the black-oat dry matter, a greater accumulation of N > K > P > Mg > Ca can be seen in Area 1, and K > N > P > Ca > Mg in Area 2; a pattern being shown for P only, N and K displaying greater variation in accumulation by the black oats in the two areas under NTS. This is probably associated with the potassium fertiliser used on previous crops (Santi et al., 2003; Nakagawa and Rosolem, 2005; Prado et al., 2006), indicating the nutrients may be released in different ways to the succeeding crop through nutrient cycling (Table 2). The results of the descriptive statistics for the NDVI generated from readings taken with the portable ground sensor and the UAV, and evaluated in both areas at two stages in the soybean (R5 and R5.5), did not show a normal distribution for most of the variables, with only NGS5 in Area 2 showing a normal distribution of the data (Table 3). Regarding the results for data dispersion, for all variables in both areas the dispersion was classified as low (CV < 15%), with coefficients ranging from 7.55 to 9.80%, and 4.35 to 10.86% for Areas 1 and 2 respectively (Table 3). Zhitao et al. (2014) also found no significant variations in NDVI in the soybean in the period from sowing to

Values for Ca were classified as moderate to high in Area 1 and high in Area 2, and values for Mg were classified as high in both areas (Commission…, 2016) (Table 1). Generally, appropriate levels of Ca and Mg in the soil are associated with a parent material (basalt) rich in minerals and the source of these nutrients (plagioclase and pyroxene) (Gergely et al., 2000), and with the use of dolomitic limestone in liming to correct soil acidity. SOM was classified as moderate in the two areas (Commission…, 2016) (Table 1). Even with the areas in question having been managed under NTS for over 20 years, the moderate levels of SOM may be related to the difficulty of increasing the content under tropical and subtropical climate conditions in soils at an advanced stage of weathering, and also to occasional errors in management (periods with no crop rotation, the correction of compaction by ploughing, among others) that did not provide an adequate supply of organic carbon to the soil. The results of the descriptive statistical analysis for the chemical composition of the dry matter from the crop of black oats, and of grain yield in the soybean crop, showed that a normal distribution was not seen for these variables in either area (Table 2). In Area 1, only the amount of Ca accumulated in the black oats showed data dispersion classified as high (CV > 35%), the other variables being classified as moderate (15–35%); in Area 2 only the accumulated N showed a dispersion classified as high (CV > 35%), the other variables being classified as having moderate dispersion (Table 3). The range of values for grain yield in Area 1 (2264.2–9717.9 kg ha−1) and Area 2 (2538.62–4885,119 kg ha−1) demonstrates that, despite being seen as a crop with characteristics of high plasticity, i.e. a high ability to adapt to environmental and management conditions through modifications in plant morphology and production components (Rambo et al., 2003), this theory is not always well accepted (Heitholt et al., 2005; Ludwig et al., 2011) where certain factors for variation in the production environment can affect yield, 612

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Table 2 Descriptive statistics for dry weight, and accumulated amounts of N, P, K, Ca and Mg in black oats, and for grain yield in soybean grown in succession. Crop Area 1 Oats

Soybean Area 2 Oats

Soybean

Variable

Mínimum

Mean

Maximum

CV (%)†

SD‡

Cs¶

Ck#

W‡‡

Dry matter (kg ha−1) Accumulated N (kg ha−1) Accumulated P (kg ha−1) Accumulated K (kg ha−1) Accumulated Ca (kg ha−1) Accumulated Mg (kg ha−1) Grain yield (kg ha−1)

2227.20 25.33 5.64 22.70 2.92 3.95 2264.20

4509.74 50.70 9.50 43.50 6.70 6.90 4856.50

7722.93 114.90 17.30 90.90 19.10 13.00 9717.90

26.37 30.57 25.37 29.84 44.48 28.84 27.50

1189.64 15.50 2.41 12.98 2.98 1.99 1335.35

0.54 0.95 0.79 0.89 2.16 0.83 0.67

−0.27 1.36 0.41 0.74 5.91 0.12 0.65

0.96* 0.94* 0.95* 0.94* 0.79* 0.93* 0.97*

Dry matter (kg ha−1) Accumulated N (kg ha−1) Accumulated P (kg ha−1) Accumulated K (kg ha−1) Accumulated Ca (kg ha−1) Accumulated Mg (kg ha−1) Grain yield (kg ha−1)

2029.47 24.86 4.70 42.35 4.83 2.79 2538.62

4136.77 51.14 9.20 92.87 8.44 5.62 4222.13

9821.73 111.72 18.47 197.75 18.41 12.86 4885.11

36.26 35.55 32.54 32.45 34.62 34.70 15.47

1500.25 18.18 2.99 30.14 2.92 1.95 653.49

1.82 1.61 1.10 1.27 1.91 2.01 −1.36

5.16 3.70 1.55 2.930 4.32 5.68 0.84

0.85* 0.86* 0.92* 0.92* 0.81* 0.82* 0.81*

†

Coefficient of variation. Standard Deviation. ¶ Coefficient of asymmetry. # Coefficient of kurtosis. ‡‡ Shapiro-Wilk test for normal distribution. * Significant at levels of p < 0.05 and ns not significant. When significant, indicates that the hypothesis of a normal distribution is rejected. ‡

3.2. Spearman’s correlation

physiological maturity, however this low variation of the NDVI does not interfere with its capacity to be a good tool for representing the variability of a crop in any given area, as the low variation in NDVI values in the study area may be related to the stage of the soybean, R5 and R5.5, i.e. with a high LAI (leaf area index), resulting in lower spectral variation. Gitelson et al. (2014) also report that compared to other annual crops, in this case maize, soybean shows less spectral variance, where this phenomenon can affect the NDVI index in different ways. It can be seen that the values for NDVI obtained with the portable sensor showed higher values than the readings taken with the UAV. These results may be associated with the fact that readings taken with a UAV tend to suffer greater influence from the soil than with the portable sensor, where measurements are made directly above the canopy with minimum influence from the soil, tending to increase NDVI values. In addition, the portable sensor has a smaller radius for the acquisition of spectral data, which makes more comprehensive measurements difficult when compared to the images obtained with the UAV.

Grain yield in the soybean showed a positive correlation with the soil chemical properties relating to P and SOM, and a negative correlation with Mg in both study areas (Tables 4 and 5). A positive correlation with the levels of P in the soil was expected due to the high dependency of the soybean crop on this nutrient, and also since tropical soils are deficient in this nutrient as a result of the parent material and the strong interaction of P with the soil where less than 0.1% can be found in the solution (Fardeau, 1996). SOM has also proved to be an important factor in soybean yield, which may be due to the beneficial effects on the chemical, physical and biological properties of the soil under NTS, providing a favourable environment for the growth and development of the soybean (Salvo et al., 2010; Mazzilli et al, 2015). The negative correlation with soil Mg can be explained by the balance in the relationship between the extraction and export of this nutrient in the soil. For the black oats, a positive correlation was primarily seen between soil P and the accumulated amounts of N, K, Ca and Mg, and also

Table 3 Descriptive statistics for the NDVI generated from readings taken with the portable ground sensor and the UAV, evaluated at the R5 and R5.5 stages in the soybean. Variable§

Minimum

Mean

Maximum

CV%†

SD‡

Cs¶

Ck#

W‡‡

Area 1 NGS5 NGS55 NR5 NR55

0.24 0.54 0.43 0.20

0.85 0.86 0.53 0.51

0.97 0.91 0.60 0.59

9.41 4.65 7.55 9.80

0.08 0.04 0.04 0.05

−4.48 −4.05 −0.43 −3.11

26.01 31.80 0.06 15.19

0.54* 0.73* 0.97* 0.74*

Area 2 NGS5 NGS55 NR5 NR55

0.68 0.40 0.29 0.22

0.79 0.48 0.47 0.48

0.86 0.51 0.52 0.54

4.35 5.28 7.58 10.86

0.03 0.02 0.03 0.05

−0.78 −1.28 −2.80 −2.15

1.32 1.84 11.10 7.45

0.95 ns 0.88* 0.74* 0.79*

§ Where: NGS5 = NDVI generated by the portable sensor at the R5 stage in the soybean; NGS5 = NDVI generated by the portable sensor at the R5.5 stage in the soybean; NR5 = NDVI generated by the UAV at the R5 stage in the soybean; NR55 = NDVI generated by the UAV at the R5.5 stage in the soybean. † Coefficient of variation. ‡ Standard deviation. ¶ Coefficient of asymmetry. # Coefficient of kurtosis. ‡‡ Shapiro-Wilk test for normal distribution. * Significant at levels of p < 0.05. ns Not significant. When significant, indicates that the hypothesis of a normal distribution is rejected.

613

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Table 4 Spearman correlation matrix between chemical properties of the soil, soybean and black oats, and readings with the portable ground sensor and with the uav, evaluated at the R5 and R5.5 stages in the soybean for Area 1. Attribute§

pH

P

K

SOM

Al

Ca

Mg

Al + H

CEC

V

m

pH P K SOM Al Ca Mg Al + H CEC V m Yld NA PA KA CaA MgA DM NGS5 NGS55 NR5 NR55

– −0.13 0.01 −0.13 0.01 0.25** 0.33** −0.46** 0.28** 0.50** −0.69** 0.03 0.01 −0.04 −0.02 0.01 0.02 −0.05 −0.01 0.11 −0.09 −0.12

– 0.10 0.03 0.04 0.07 0.18* −0.11 0.38** 0.13 −0.06 0.34** 0.37** 0.28** 0.41** 0.29** 0.42** 0.38** −0.13 0.28** 0.01 −0.05

– 0.15 0.03 −0.01 0.05 −0.08 −0.09 0.05 0.08 0.14 0.01 −0.03 0.06 0.15 0.00 0.02 0.07 0.01 0.01 −0.13

– 0.07 0.08 −0.01 0.24** 0.27** 0.37* 0.15 0.27* −0.05 −0.02 0.07 0.09 0.02 0.01 0.35** 0.03 0.25** 0.15

– −0.10 0.05 0.08 0.09 −0.12 0.06 −0.05 −0.06 −0.05 −0.03 0.01 −0.00 −0.02 0.02 0.01 0.07 −0.02

– 0.48** −0.20* 0.33** 0.52** −0.42** −0.12 −0.04 −0.05 −0.01 −0.05 0.01 −0.06 −0.01 −0.06 −0.13 −0.04

– −0.24** 0.15 0.46** −0.36** −0.20* −0.00 0.02 −0.03 −0.00 0.02 −0.06 −0.07 −0.02 0.14 −0.12

– −0.75** 0.87** −0.52** 0.10 −0.18* −0.03 −0.24** −0.12 −0.17* −0.12 −0.10 0.07 0.08 0.15

– 0.45** −0.22** 0.02 0.20* −0.05 0.25** −0.15 −0.14 0.15 −0.12 0.02 −0.04 0.06

– −0.63** −0.10 0.13 0.05 0.18* 0.08 0.15 0.09 0.08 −0.07 −0.13 −0.15

– −0.06 −0.11 −0.09 −0.06 −0.02 −0.14 −0.09 0.05 −0.09 0.06 0.07

Attribute

Yld

NA

PA

KA

CaA

MgA

DM

NGS5

NGS55

NR5

NR55

Yld NA PA KA CaA MgA DM NGS5 NGS55 NR5 NR55

– 0.30* 0.22* 0.14 0.06 0.26** 0.23** 0.20* 0.31** 0.47** 0.27**

– 0.89** 0.71** 0.80** 0.94** 0.92** 0.14 0.13 0.29* 0.08

– 0.60** 0.75** 0.89** 0.87** 0.01 0.20* 0.26* 0.06

– 0.60** 0.76** 0.78** 0.33** 0.19* 0.12 0.03

– 0.82** 0.71** 0.04 0.20* 0.09 −0.01

– 0.91** 0.15 0.21* 0.19* 0.03

– 0.20* 0.18* 0.26** 0.23**

– 0.01 0.45** 0.01

– 0.01 0.40**

– 0.15

–

§ Where: P = phosphorus; K = potassium; SOM = soil organic matter; Al = exchangeable acidity; Ca = calcium; Mg = magnesium; H + Al = acidity potential; CEC = cation exchange capacity; V = base saturation; m = aluminum saturation; Yld = soybean grain yield; DM = black-oat dry matter; NA = nitrogen accumulated in the black-oat dry matter; PA = phosphorus accumulated in the black-oat dry matter; KA = potassium accumulated in the black-oat dry matter; CaA = calcium accumulated in the black-oat dry matter; MgA = magnesium accumulated in the black-oat dry matter; NGS5 = NDVI generated with the portable sensor at the R5 stage in the soybean; NGS55 = NDVI generated with the portable sensor at the R5.5 stage in the soybean; NR5 = NDVI generated with the UAV at the R5 stage in the soybean; NR55 = NDVI generated with the UAV at the R5.5 stage in the soybean. * Significant at 5% probability. ** Significant at 1% probability.

The NDVI readings taken with the portable ground sensor and also with the UAV showed a positive correlation for NGS55 with the levels of soil P, and for NGS5 and NR5 with the SOM for the two areas (Tables 4 and 5). There was also a correlation with yield for the two forms of evaluating the NDVI at both stages of the soybean; the greater part of these correlations were significant at 1% probability of error, showing that the NDVI can be highly reliable in estimating the grain yield of the soybean crop. NGS5, NGS55, NR5 and NR55 displayed a positive correlation with dry matter, and NGS55 and NR5 displayed a positive correlation with the accumulated amounts of P and Mg, in the two areas under study. In the case of NGS55 and NR5, there was a correlation with the accumulated amounts of N and Ca respectively, whereas NGS5 and NGS55 correlated with the accumulated amount of K (Tables 4 and 5). As expected, the highest correlations found for the NDVI are seen for cycled nutrients rather than directly by the effect of the black-oat dry matter. In general, it can be seen that there is a tendency for the nutrients cycled by the black oats and subsequently evaluated by the NDVI in the soybean to vary according to the soybean stage and forms of evaluation; this may be due to the dynamics of nutrient availability in the soil and the physiological responses of the soybean. It should be noted that positive correlations were seen between the readings made with the portable ground sensor and with the UAV, both at the same stages and at different stages for the two areas of study.

the DM in both study areas (Tables 4 and 5). This effect is due to P having an influence on most of the macronutrients exported in the DM of the black oats (Melo et al, 2011), where in this case, the high levels in the soil favoured an increase in DM. However as discussed above, in this study, the P content of the soil showed considerable variation in the study area, which probably influenced nutrient accumulation in the black-oat DM. Furthermore, there was a significant correlation between the accumulated amounts of N and K with soil CEC, in addition to the negative correlation between the accumulated quantities of K and Mg with AL + H, which confirms the beneficial effect of increasing the CEC, and on the other hand, the opposite effect of potential acidity in tropical soils cultivated under NTS. The dry matter and accumulated amounts of N, P and Mg in the oats displayed a positive correlation with grain yield in the soybean in both study areas (Tables 4 and 5). According to Crusciol et al. (2008), in studies into rates of decomposition and macronutrient release in straw from black oats in a Red Latosol under NTS, they found that the points of maximum accumulated release were reached at 63 and 50 days after harvesting, for N and Mg respectively, whereas for P, the authors failed to reach a point of maximum release, which generally characterises a gradual release, coinciding with the R3/R4 stage of the soybean in this study, confirming the positive effects on grain yield. For the black-oat dry matter, the same authors found that at 53 days after harvest only 33.6% of the initial amount of dry matter remained. 614

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Table 5 Spearman correlation matrix between chemical properties of the soil, soybean and black oats, and readings with the portable ground sensor and with the UAV, evaluated at the R5 and R5.5 stages in the soybean for Area 2. Attribute§

pH

P

K

SOM

Al

Ca

Mg

Al + H

CEC

V

m

pH P K SOM Al Ca Mg Al + H CEC V m Yld NA PA KA CaA MgA DM NGS5 NGS55 NR5 NR55

– −0.18 0.16 0.07 0.19 0.36* 0.69** −0.62** 0.45* 0.36* 0.02 0.15 0.05 0.10 −0.13 −0.19 −0.17 −0.14 0.06 0.11 0.08 0.02

– 0.17 −0.33* 0.10 0.18 0.32* −0.03 0.31* −0.22 −0.12 0.39* 0.35* 0.08 0.22* 0.24* 0.27* 0.39** 0.21* −0.17 −0.12 0.13

– −0.18 0.27 0.40* −0.02 0.18 0.02 −0.19 −0.29 0.11 −0.10 0.04 −0.19 −0.16 −0.15 −0.09 0.27 0.36* 0.33* 0.30

– −0.12 0.03 0.05 0.32* 0.33* 0.88** −0.15 0.37* 0.30 −0.10 0.29 0.20 0.26 0.28 0.27* 0.35* 0.40** 0.23*

–− −0.25* −0.28* 0.29** −0.30** −0.31** −0.98** −0.01 0.21 −0.05 0.15 0.09 0.15 −0.17 0.17 0.15 −0.03 0.02

– 0.89** −0.13 0.51** 0.25* −0.78** 0.17 −0.16 0.05 −0.17 −0.09 −0.11 −0.14 0.04 0.10 0.12 0.16

– −0.01 0.59** 0.08 −0.82** −0.27* −0.14 0.09 −0.28 −0.05 −0.14 −0.17 0.02 0.05 0.16 0.19

– 0.15 0.65** −0.25 0.01 0.07 −0.11 −0.28* 0.02 −0.25* −0.03 0.15 0.11 0.08 0.05

– 0.66** −0.33** 0.05 0.27* 0.04 0.24* 0.15 −0.12 −0.13 0.03 0.01 0.12 0.06

– −0.30* 0.02 −0.08 0.13 0.06 −0.15 0.05 0.03 −0.13 −0.06 0.08 0.13

– −0.04 0.20 −0.08 0.10 0.09 0.15 0.17 0.15 0.13 0.07 0.02

Attribute

Yld

NA

PA

KA

CaA

MgA

DM

NGS5

NGS55

NR5

NR55

Yld NA PA KA CaA MgA DM NGS5 NGS55 NR5 NR55

– 0.31** 0.34** 0.10 0.12 0.27* 0.33** 0.35** 0. 47** 0.73** 0.53**

– 0.68** 0.85** 0.88** 0.82** 0.83** 0.12 0.29* 0.16 0.21*

– 0.67** 0.64** 0.72** 0.64** 0.29* 0.27* 0.39** 0.25*

– 0.82** 0.85** 0.83** 0.23* 0.21* 0.15 0.06

– 0.89** 0.74** 0.03 0.22* 0.05 0.25*

– 0.82** 0.04 0.03 0.01 0.18

– 0.38** 0.15 0.33** 0.25*

– 0.57** 0.49** 0.36*

– 0.31** 0.48**

– 0.12

–

§ Where: P = phosphorus; K = potassium; SOM = soil organic matter; Al = exchangeable acidity; Ca = calcium; Mg = magnesium; H + Al = acidity potential; CEC = cation exchange capacity; V = base saturation; m = aluminum saturation; Yld = soybean grain yield; DM = black-oat dry matter; NA = nitrogen accumulated in the black-oat dry matter; PA = phosphorus accumulated in the black-oat dry matter; KA = potassium accumulated in the black-oat dry matter; CaA = calcium accumulated in the black-oat dry matter; MgA = magnesium accumulated in the black-oat dry matter; NGS5 = NDVI generated with the portable sensor at the R5 stage in the soybean; NGS55 = NDVI generated with the portable sensor at the R5.5 stage in the soybean; NR5 = NDVI generated with the UAV at the R5 stage in the soybean; NR55 = NDVI generated with the UAV at the R5.5 stage in the soybean. * Significant at 5% probability. ** Significant at 1% probability.

NR55 and NR5 showed a positive correlation with NGS5 and NGS55, and NR55 even showed a positive correlation with NR5. These relationships, especially between readings taken with the portable ground sensor and those taken with the UAV, are noteworthy, as there remained uncertainties regarding NDVI readings taken with a UAV, since this technology is still new and requires validation as to its use in agriculture. As seen in this study, readings taken with the UAV are positively correlated with readings taken with the portable ground sensor, so that for evaluations using the NDVI, the portable ground sensor can be replaced by the UAV, primarily for ease of operation in large areas, as is the case in this study (Tables 4 and 5). From the parameters under evaluation relating to the chemical properties of the soil, the nutrients accumulated in the black-oat dry matter, and the NDVI readings at the R5 and R5.5 stages in the soybean taken with both the portable ground sensor and with the UAV, those that showed a correlation with grain yield in the soybean were selected by adjusting the regression models using stepwise equations, where only those parameters that actually influenced grain yield were included in the equation models. The results showed three and two significant equations for Areas 1 and 2 respectively (Table 6). The first equation for Area 1 included the black-oat dry matter only, but had the lowest coefficient of determination (r2) among the three equations. In the second equation, the variables included the SOM and the N accumulated in the black oats, with an r2 of 0.66. The third

Table 6 Stepwise equation for estimating dependent variable components for grain yield in the soybean. Study area

Equation

Equation parameters§

r2

SE†

DW‡

Area 1

1 2

Y = 3787.55 + 0.30 (DM) Y = 2686.81 + 16.44(NA) + 436.91(SOM) Y = -6655.33 + 16230.65(NR5) + 3134.94(NGS5) + 100.25(MgA)

0.55 0.66

28 38

1.50 1.45

0.75

24

1.66

0.63

13

1.55

0.70

11

1.47

3 Area 2

1 2

Y = -1435.89 + 1404.64 (SOM) + 58.40 (PA) Y = -5264 + 8807.87(NR5) + 4613.69(NR55) + 3526.02 (NGS5) + 5.21 (NA)

§ Y = soybean grain yield; DM = black-oat dry matter; NA = nitrogen accumulated in the black oats; PA = phosphorus accumulated in the black-oat dry matter; SOM = organic matter; NR5 = NDVI generated with the UAV at the R5 stage in the soybean; NR55 = NDVI generated with the UAV at the R5.5 stage in the soybean; NGS5 = NDVI generated with the portable sensor at the R5 stage in the soybean; MgA = magnesium accumulated in the black oats. † SE: Residual standard error (%). ‡ DW: Durbin-Watson test.

615

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

and soybean grain yield, and the spherical model to N accumulated in the oats, NR5 and NGS5. In Area 2, the greatest ranges were seen for the accumulated P and N, with values of 592 and 591 m respectively, with the remaining variables in descending order, NR55, NR5, soybean grain yield, SOM and NGS5, with values of 467, 450, 404, 359 and 304 m respectively. The exponential model was fitted to the accumulated P, accumulated N, SOM and soybean grain yield, and the spherical model to the NR5, NR55 and NGS5 readings. Spatial dependence was classified as moderate for accumulated P, accumulated N, SOM and NR55, and as strong only for NR5, NGS5 and soybean grain yield.

equation consisted of NR5, NGS5 and the amount of Mg accumulated in the black oats, with an r2 of 0.85, the highest among the three equations for Area 1. For Area 2, equation 1 consisted of the SOM and P accumulated in the black oats, with an r2 of 0.63, and equation 2 consisted of the variables NR5, NR55, NGS5 and accumulated N, with an r2 of 0.70, the greater of the two equations for Area 2. The residual standard error was similar and considered as low in the three equations for Area 1, and in the two equations for Area 2, with a range of from 24% to 28% and from 11% to 13% respectively. As with the standard error, the Durbin-Watson test was similar in the three equations for Area 1, and in the two equations for Area 2, with values ranging from 1.45 to 1.66 and from 1.47 to 1.55, close to the ideal value of 2, showing that the data did not exhibit autocorrelation (Neter et al., 1985). It can be seen that among the soil parameters under evaluation, only the SOM integrated the equations for the two areas, again showing that one of the main obstacles to production systems under NTS in tropical soils is the SOM. Of the nutrients accumulated by the black oats, only the N was maintained in the equations for the two areas, with the response to the accumulated P and Mg showing the greatest differences as to the conditions of each production system. In addition, the influence of the NDVI readings taken with the portable ground sensor and with the UAV is noteworthy, these variables being included in the equations for the two areas under study. As it was not possible to obtain an equation that would uniquely explain grain yield in the soybean, it was decided that all the variables that were fitted to the three equations for Area 1, and the two equations for Area 2, were of importance to soybean yield, these variables being used in the remaining evaluations of this study.

3.4. Spatial distribution maps After the geostatistical analysis, it was possible to carry out the spatial distribution of values for the variables in Areas 1 and 2 (Figs. 4 and 5). For Area 1, based on the map for soybean grain yield (Fig. 4 g), it can be seen that the sub-areas with the highest yield, corresponding to a band that runs from south to east on the map, coincided, especially at higher levels, with the maps for dry matter (Fig. 4a), accumulated N (Fig. 4b), SOM (Fig. 4c), NR5 (Fig. 4d), NGS5 (Fig. 4e) and accumulated Mg (Fig. 4f), again reinforcing the importance of these variables for grain yield in the soybean. It should be noted that the map for black-oat dry matter showed the highest degree of similarity among the maps discussed above, with this feature having a direct influence on accumulated N and accumulated Mg, in addition to SOM in the long term. In Area 2, on the map for soybean grain yield (Fig. 5 g), an area of higher yield was found, extending from southeast to northwest, which is also present on the maps for SOM (Fig. 5c), NR5 (Fig. 5d), NR55 (Fig. 5e) and NGS5 (Fig. 5f). The maps for accumulated P (Fig. 5a) and accumulated N (Fig. 5b) also showed correspondence, mainly to areas of higher yield, on the map for soybean grain yield, but in smaller areas showing less correspondence in the sub-areas located mainly to the southeast and southwest of the map, which again confirms the influence of the nutrients cycled by the black oats on grain yield.

3.3. Geostatistical analysis From the results of the geostatistical analysis with variables selected by the stepwise equation for Areas 1 and 2, it was found that all variables presented a structure for the semivariogram, allowing interpretation and projection of the results based on the structure of their variability (Table 7). For Area 1, the greatest ranges were seen for Mg accumulated in the black oats, and for grain yield in the soybean, of 989 and 830 m respectively; values for the remaining variables in descending order were 797, 611, 417, 395 and 156 m, for accumulated N, NGS5, black-oat dry matter, NR5 and SOM respectively (Table 7). The exponential model was fitted to the SOM, dry matter, accumulated Mg in the black oats,

3.5. Definition of management zones To define management zones using the fuzzy c-means algorithm, initially the variables relating to dry matter, accumulated N, SOM NR5, NGS5 and accumulated Mg for Area 1, and accumulated P, accumulated N, SOM, NR5, NR55 and NGS5 for Area 2, were submitted to testing using the fuzzy performance index (FPI) and normalized classification entropy index (NCE), seeking to determine the optimal number of

Table 7 Geostatistical analysis of the variables selected by stepwise equation and for grain yield in the soybean. Variable§

Nugget effect (C0)

Sill (C0 + C1)

Contribution (C1)

Range (a)

Model

r2

Area 1 DM (kg ha−1) NA (kg ha−1) SOM (g dm−3) NR5 NGS5 MgA (kg ha−1) Yld (kg ha−1)

969000 64.60 0.04 0.0001 0.0003 2.07 1085880

1939000 182.60 0.21 0.0007 0.0008 4.15 2214680

970000 118 0.17 0.0006 0.0005 2.08 1128800

417 797 156 395 611 989 830

Exponential Spherical Exponential Spherical Spherical Exponential Exponential

0.98 0.97 0.81 0.86 0.95 0.88 0.73

Area 2 PA (kg ha−1) NA (kg ha−1) SOM (g dm−3) NR5 NR55 NGS5 Yld (kg ha−1)

357.46 92.95 0.02 0.0001 0.001 0.0003 23000

607.55 138.32 0.08 0.0007 0.003 0.002 1072561

250.55 45.37 0.06 0.0006 0.002 0.002 1049561

592 591 359 450 467 304 404

Exponential Exponential Exponential Spherical Spherical Spherical Exponential

0.86 0.83 0.89 0.95 0.87 0.84 0.96

§ DM = black-oat dry matter; NA = nitrogen accumulated in the black oats; PA = phosphorus accumulated in the black-oat dry matter; MgA = magnesium accumulated in the black oats; SOM = organic matter; NR5 = NDVI generated with the UAV at the R5 stage in the soybean; NR55 = NDVI generated with the UAV at the R5.5 stage in the soybean; NGS5 = NDVI generated with the portable sensor at the R5 stage in the soybean; Yld = soybean grain yield.

616

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Fig. 4. Spatial distribution maps for black-oat dry matter (kg ha−1) (a), accumulated N (kg ha−1) (b), SOM (c) NR5 (d) NGS5 (e) accumulated Mg (kg ha−1) (f) and soybean grain yield (kg ha−1) (g) for Area 1.

be seen in the clustering process. Based on this, the results show that three zones would be the ideal number to represent the set of variables relating to dry matter, accumulated N, SOM, NR5, NGS5 and accumulated Mg in Area 1 (Fig. 7a), and two zones to represent the set of variables relating to accumulated P, accumulated N, SOM NR5, NR55 and NGS5 in Area 2 (Fig. 7b). It can be seen that the maps of the management zones defined for the two areas (Fig. 7) displayed a great similarity with the maps of the

classes for setting up the management zones (Fig. 6). Based on the above, it can be seen that in the two areas, as the number of management zones increased, there was generally a rise in the values of the two indices, signifying an increase in error when indicating the zones. According to the criterion adopted by Fridgen et al. (2004), the lowest values for the NCE and FPI result in a suitable number of management zones, since according to those authors, the least associated division (FPI) or the greatest amount of organization (NCE) can

617

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Fig. 5. Spatial distribution maps for accumulated P (kg ha−1) (a), accumulated N (kg ha−1) (b), SOM (c), NR5 (d), NR55 (e), NGS5 (f) and soybean grain yield (kg ha−1) (g) for Area 2.

in Zone 3, the adjustment could be made through the foliar application of 38 and 17.13 kg ha−1 as magnesium sulphate (16% Mg) in Zones 1 and 2 respectively. In Area 2, in the case of P, based on the average value of 10.62 kg ha−1 P cycled by the black oats in Zone 2, an adjustment could be made by the application of 10.46 kg ha−1 in the form of triple superphosphate (45% P) in Zone 1, and in the case of N, based on the average value of 88.72 kg ha−1 N cycled by the black oats in Zone 2, an adjustment could be made by the application of 112.04 kg ha−1 as urea (45% N) in Zone 1. In the long term, in the case of dry matter in Area 1, other plants such as the turnip could be used in a winter rotation scheme, or even

variables initially defined for Areas 1 (Fig. 4) and 2 (Fig. 5), and that from the values found with these maps, it can be assumed that on the maps of the management zones, especially in Zone 1 to Zone 2 in Area 1, and in Zone 1 in Area 2, it should possible to intervene in the soybean crop in the short and long term. In the short term, measures could be taken when carrying out varied and/or fixed rate fertilisation where, for example in Area 1 in the case of N, and based on the average value of 81.90 kg ha−1 N cycled by the black oats in Zone 3, an adjustment could be made by the application of 97.84 and 59.8 kg ha−1 in the form of urea (45% N) in Zones 1 and 2 respectively, and in the case of Mg, based on the mean value of 10.07 kg ha−1 Mg cycled by the black oats 618

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Fig. 6. Fuzzy performance index (FPI) and the normalized classification entropy index (NCE) for Areas 1 (a) and 2 (b).

Fig. 7. Map of the management zones using the fuzzy c-means algorithm method of clustering for Areas 1 (a) and 2 (b).

4. Conclusions

intercropped with the black oats, so that the variability caused in the winter would be reduced. From a comparison of the mean values for dry mass, accumulated N, SOM, NR5, NGS5 and accumulated Mg in the three management zones for Area 1, and between the variables for accumulated P, accumulated N, SOM, NR5, NR55 and NGS5 for Area 2, it was found that the method was generally effective in diagnosing the differences between management zones for each variable (Table 8). By analysing the results, management Zones 3 and 1 in Area 1 were defined as of high and low potential respectively, whereas in Area 2, Zones 2 and 1 were defined as of high and low potential respectively. Only for SOM was there no difference between management Zones 1 and 2 in Area 1, and between the two zones defined in Area 2 (Table 8). The absence of a statistical difference could be linked to the low variation of this variable in the soil; however, as low as these differences may be, the variable should be monitored, as such differences can influence various processes in the soil, where the sum of these factors would result in a significant effect on the soil system.

It was found that variability in the cover crop has an influence on yield in the cash crop, where in the case of this study, the black oats influenced the yield of the soybean grain, demonstrating that this factor should be taken into account in fertilisation and crop management in areas cultivated under a no-tillage system. Among the main factors involved are the amount of dry matter produced by the black-oat crop, as well as the accumulated nutrients relating to nitrogen, phosphorus and magnesium, where alternatives such as adjustments to the base and/or cover fertiliser, as well as managing the system of crop rotation, may help to correct this variation in the main crop, in this case, the soybean. For the soil chemical properties under evaluation, organic matter had the most influence on soybean yield, showing it to be a factor that should be taken into consideration regarding variability in areas under NTS, seeking to maintain the appropriate levels to preserve the sustainability of the soybean production system. The NDVI evaluated with the portable ground sensor and with the UAV, were both efficient in assessing the effects of variability in the 619

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

Table 8 Comparison of mean values for the variables in the three management zones defined with the fuzzy c-means algorithm method of clustering. Zone

Zone 1 Zone 2 Zone 3

Zone 1 Zone 2

Cerri, C., Maia, S.M.F., Galdos, M.V., Cerri, C.E.P., Feigl, B.J., Bernoux, M., 2009. Brazilian greenhouse gas emissions: the importance of agriculture. Sci. Agric. 66, 831–843. Commission for Soil Chemistry and Fertility – RS/SC. Manual of liming and fertilization for the states of Rio Grande do Sul and Santa Catarina. Commission of soil chemistry and fertility, Porto Alegre, RS, Brazil. (in Portuguese). Costa, M.S.S.M., Pivetta, L.A., Costa, L.A.M., Pivetta, L.G., Castoldi, G., Steiner, F., 2011. Soil physical attributes and corn yield as affected by soil managements and fertilization. Rev. Bras. Eng. Agríc. Ambient. 15, 810–815. Crusciol, C.A., Soratto, R.P., 2009. Nitrogen supply for cover crops and effects on peanut grown in succession under a no-till system. Agron. J. 101, 41–46. Crusciol, C.A.C., Moro, E., Lima, E.V., Andreotti, M., 2008. Decomposition rate and nutrient release of oat straw used as mulching in no-till system. Bragantia 67, 481–489. Cunha, E.Q., Stone, L.F., Ferreira, E.P.B., Didonet, A.D., Moreira, J.A.A., 2012. Physical, chemical and biological attributes of soil under organic production as impacted by crop systems. Rev. Bras. Eng. Agríc. Ambient. 16, 56–63. Dalchiavon, F.C., Carvalho, M.P., Andreotti, M., Montanari, R., 2012. Spatial variability of the fertility attributes of Dystropheric Red Latosol under a no-tillage system. Rev. Cienc. Agron. 43, 453–461. Damian, J.M., Pias, O.H.C., Santi, A.L., Di Virgilio, N., Berghetti, J., Barbanti, L., Roberta Martelli, R., 2016. Delineating management zones for precision agriculture applications: a case study on wheat in sub-tropical Brazil. Ital. J. Agron. 11, 171–179. Da Ros, C.O., Aita, C., 1996. Cover crops as a source of nitrogen to corn. Rev. Bras. Cienc. Solo 20, 135–140 (in Portuguese). Dugo, V.G., Zarco-Tejada, P., Nicolás, E., Nortes, P.A., Alarcón, J.J., Intrigliolo, D.S., Fereres, E., 2013. Using high resolution UAV thermal imagery to assess the variability in the water status of five fruit tree species within a commercial orchard. Precis. Agric. 14, 660–678. Fardeau, J.C., 1996. Dynamics of phosphate in soils: an isotopic outlook. Fert. Res. 45, 91–100. Ferraz, G.A.E.S., Silva, F.M., Carvalho, L.C.C., Alves, M.C., Franco, B.C., 2012. Variabilidade espacial e temporal do fósforo, potássio e da produtividade de uma lavoura cafeeira. Eng. Agríc. 32, 140–150. Fridgen, J.J., Kitchen, N.R., Sudduth, K.A., Drummond, S.T., Wiebold, W.J., Fraisse, C.W., 2004. Management zone analyst (MZA): software for subfield management zone delineation. Agron. J. 96, 100–108. Gergely, A.J., Babinski, M., Teixeira, W., 2000. Igneous rocks. In: Teixeira, W. (Ed.), Deciphering the Earth. Oficina de Textos, São Paulo, SP, pp. 327–346. Giotto, E., Robaina, A.D., 2007. Precision Farming With the CR Campeiro 7. Federal University of Santa Maria, Santa Maria, RS, Brazil (in Portuguese). Gitelson, A.A., Peng, Y., Huemmrich, K.F., 2014. Relationship between fraction of radiation absorbed by photosynthesizing maize and soybean canopies and NDVI from remotely sensed data taken at close range and from MODIS 250 m resolution data. Remote Sens. Environ. 147, 108–120. Gonçalves, C.N., Ceretta, C.A., Basso, C.J., 2000. Plant successions of cover crops and corn under no tillage system and their effects on soil nitrogen. Rev. Bras. Cienc. Solo 24, 153–159 (in Portuguese). Granados, F.L., Torres-Sánchez, J., Serrano-Pérez, A., Castro, A.I., Mesas-Carrascosa, F.J., Peña, J.M., 2016. Early season weed mapping in sunflower using UAV technology: variability of herbicide treatment maps against weed thresholds. Precis. Agric. 17, 183–199. Heitholt, J.J., Farr, J.B., Eason, R., 2005. Planting configuration × cultivar effects on soybean production in low-yield environments. Crop Sci. 45, 1800–1808. Jones, J.R., Fleming, C.S., Pavuluri, K., Alley, M.M., Reiter, M.S., Thomason, W.E., 2015. Influence of soil, crop residue, and sensor orientations on NDVI readings. Precis. Agric. 16, 690–704. Kondo, M.K., Albuquerque, C.J.B., Wendling, B., Silva, P.B., Cardoso, M.M., 2012. Effect of cover crops on soil physical attributes and agronomic characteristics of grain sorghum. Biosci. J. 28, 33–40. Lopresti, M.F., Di Bella, C.M., Degioanni, A.J., 2015. Relationship between MODIS-NDVI data and wheat yield: a case study in Northern Buenos Aires province, Argentina. Inf. Process. Agric. 2, 73–84. Ludwig, M.P., Dutra, L.M.C., Filho, O.A.L., Zabot, L., Jauer, A., Uhry, D., 2011. Soybean plant stand in conventional cultivars and Roundup Ready™. Rev. Ceres 58, 305–313. Mazzilli, S.R., Kemanian, A.R., Ernsta, O.R., Jackson, R.B., Piñeiro, G., 2015. Greater humification of belowground than aboveground biomass carbon into particulate soil organic matter in no-till corn and soybean crops. Soil Biol. Biochem. 85, 22–30. Melo, A.V., Galvão, J.C.C., Braun, H., Santos, M.M., Coimbra, R.R., Silva, R.R., Reis, W.F., 2011. Nutrient uptake and biomass production of black oats in soils under eighteen years of organic and mineral fertilizer. Semin: Cien. Agrar. 32, 411–420. Nanni, M.R., Povh, F.P., Demattê, J.A.M., Oliveira, R.B., Chicati, M.L., Cezar, E., 2011. Optimum size in grid soil sampling for variable rate application in site-specific management. Sci. Agric. 68 (386–392), 2011. Nakagawa, J., Rosolem, C.A., 2005. Nutrient concentration in leaves and grains of black oat affected by phosphorus and potassium fertilization. Bragantia 64, 441–445. Neter, J., Wasserman, W., Kutner, M.H., 1985. Applied Linear Statistical Methods. Irwin Professional Publishing, Illinois, IL, USA, pp. 50–78. Odeh, I.O.A., Chittleborough, D.J., Mcbratney, A.B., 1992. Soil pattern recognition with fuzzy-c-means: application to classification and soil-landform interrelationships. Soil Sci. Soc. Am. J. 56, 505–516. Passari, L.M.G.Z., Soares, P.K., Bruns, R.E., Scarminio, I.S., 2011. Statistics applied to chemistry: ten common doubts. Quim. Nova 34, 888–892. Peralta, N.R., Assefa, Y., Du, J., Barden, C.J., Ciampitti, I.A., 2016. Mid-season high-resolution satellite imagery for forecasting site-specific corn yield. Remote Sens. 8, 1–16. Prado, R.M., Romualdo, L.M., Vale, D.W., 2006. Black oats’ response to the application

Variable§ DM (kg ha−1)

NA (kg ha−1)

MgA (kg ha−1)

SOM (g dm−3)

NGS5

NR5

Yld (kg ha−1)

3693.65 c† 5406.36 b 6251.05 a

37.87 c

3.99 c

3.54 b

0.32 c

0.44c

3.335.56 c

54.99 b

7.33 b

3.62 b

0.50 b

0.51b

5.268.65 b

81.90 a

10.07 a

4.13 a

0.88 a

0.60 a

8.088.24 a

PA (kg ha−1)

NA (kg ha−1)

SOM (g dm−3)

NR5

NR55

NGS5

Yld (kg ha−1)

5.91b 10.62a

38.30b 88.72a

3.55a 3.71a

0.42b 0.57a

0.29b 0.53a

0.71b 0.84a

2705.00b 4754.20a

§ DM = black-oat dry matter; NA = nitrogen accumulated in the black oats; PA = phosphorus accumulated in the black-oat dry matter; MgA = magnesium accumulated in the black oats; SOM = organic matter; NR5 = NDVI generated with the UAV at the R5 stage in the soybean; NR55 = NDVI generated with the UAV at the R5.5 stage in the soybean; NGS5 = NDVI generated with the portable sensor at the R5 stage in the soybean; Yld = soybean grain yield. † Mean values followed by same letter do not differ statistically by Tukey’s test at 5% probability (p ≤ 0.05).

black-oat crop, the corresponding dry matter produced, and the accumulated nutrients of phosphorus, nitrogen and magnesium on grain yield in the soybean. These results show that the NDVI can be an important tool, aiding the decision-making process and controlling variability in areas under NTS. Acknowledgements The authors wish to thank The Coordination for the Improvement of Higher Education Personnel (CAPES) for the Master's scholarship. References Aita, C., Basso, C.J., Ceretta, C.A., Gonçalves, C.N., Da Ros, C.O., 2001. Cover crops as a source of nitrogen to corn. Rev. Bras. Cienc. Solo 25, 157–165. Amado, T.J.C., Bayer, C., Eltz, F.L.F., Brum, A.C.R., 2001. Potential of cover crops to sequester carbon and increase soil nitrogen content, under no-tillage system, improving environmental quality. Rev. Bras. Cienc. Solo 25, 189–197. Anjos, L.H.C., Jacomine, P.K., Santos, H.G., Oliveira, V.A., Oliveira, J.B., 2012. Brazilian system of soil classification. In: Ker, J.C. (Ed.), Pedology: Fundamentals. Brazilian Society of Soil Science, Viçosa, MG, Brazil, pp. 303–343 (in Portuguese). Alvares, C.A., Stape, J.L., Sentelhas, P.C., Moraes, J.L., Gerd, G., 2013. Köppen’s climate classification map for Brazil. Met. Zeit. 22, 711–728. Bellvert, J., Zargo-Tejada, P.J., Girona, J., Fereres, E., 2014. Mapping crop water stress index in a ‘Pinot-noir’ vineyard: comparing ground measurements with thermal remote sensing imagery from an unmanned aerial vehicle. Precis. Agric. 15, 361–376. Bezdek, J.C., 1981. Pattern Recognition With Fuzzy Objective Function Algorithms. Academic Publishers, Norwell, MC, USA, pp. 43–93. Boer, C.A., Assis, R.L., Silva, G.P., Braz, A.J.B.P., Barroso, A.L., Filho, A.C., Pires, F.R., 2007. Nutrient cycling in off-season cover crops on a Brazilian savanna soil. Pesq. Agropec. Bras. 42, 1269–1276. Boken, V.K., Shaykewich, C.F., 2002. Improving an operational wheat yield model using phenological phase-based Normalized Difference Vegetation Index. Int. J. Rem. Sens. 23, 4155–4168. Bressan, S.B., Nóbrega, J.C.A., Nóbrega, R.S.A., Barbosa, R.S., Sousa, L.B., 2013. Plant cover and chemical quality of Oxisol under no-tillage in the cerrado of Maranhão State. Rev. Bras. Eng. Agríc. Ambient. 17, 371–378. Calderón, R., Montes-Borrego, M., Landa, B.B., Navas-Cortés, J.A., Zarco-Tejada, P.J., 2014. Detection of downy mildew of opium poppy using high-resolution multispectral and thermal imagery acquired with an unmanned aerial vehicle. Precis. Agric. 15, 369–661. Carvalho, A.M., Marcão, R.L., Souza, K.W., Bustamante, M.M.C., 2014. Soil fertility status, carbon and nitrogen stocks under cover crops and tillage regimes. Rev. Ciênc. Agron. 45, 914–921. Cavalcante, E.G.S., Alves, M.C., Souza, Z.M., Pereira, G.T., 2007. Spatial variability of chemical attributes of a soil under different uses and managements. Rev. Bras. Cienc. Solo 31, 1329–1339.

620

Computers and Electronics in Agriculture 142 (2017) 607–621

J.M. Damian et al.

AVHRR, SPOT-VGT, and MODIS global NDVI products. Remote Sens. Environ. 112, 535–549. Teixeira, E.I., Johnstone, P., Chakwizira, E., Ruiter, J., Malcolm, B., Shaw, N., Zyskowski, R., Khaembah, E., Sharp, J., Meenken, E., Fraser, P., 2016. Sources of variability in the effectiveness of winter cover crops for mitigating N leaching. Agr. Ecosyst. Environ. 220, 226–235. Tedesco, M.J., Gianello, C., Bissani, C.A., Bohnen, H., Volkweiss, S.J., 1995. Analysis of Soil, Plants and Other Materials. Federal University of Rio Grande do Sul, Porto Alegre, PA, Brazil, pp. 83–114 (in Portuguese). Torres, J.L.R., Pereira, M.G., Fabian, A.J., 2008. Cover crops biomass production and its residues mineralization in a Brazilian no-till Oxisol. Pesq. Agropec. Bras. 43, 421–428. Webster, R., Oliver, M.A., 2007. Geostatistics for Environmental Scientists. John Wiley & Sons Ltd., West Sussex, WS, England, pp. 139–266. Wilding, L.P., Drees, L.R., 1983. Spatial variability and pedology. In: Wilding, L.P. (Ed.), Pedogenesis and Soil Taxonomy – I. Concepts and Interactions. Elsevier, Amsterdam, AM, Netherlands, pp. 83–116. Zhitao, Z., Lan, Y., Pute, W., Wenting, H., 2014. Model of soybean NDVI change based on time series. Int. J. Agric. & Biol. Eng. 7, 64–70.

phosphorus under two levels of nitrogen under greenhouse conditions. Acta Sci. Agron. 28, 527–533. Rambo, L., Costa, J.A., Pires, J.L.F., Parcianello, G., Ferreira, F.G., 2003. Soybean yield response to plant arrangement. Cienc. Rural 33, 405–411. Santos, G.G., Silveira, P.M., Marchão, R.L., Becquer, T., Balbino, L.C., 2008. Soil macrofauna communities and cover crops in a Cerrado Oxisol under no tillage. Pesq. agropec. bras. 43, 115–122. Salvo, L., Hernandez, J., Ernst, O., 2010. Distribution of soil organic carbon in different size fractions, under pasture and crop rotations with conventional tillage and no-till systems. Soil Till. Res. 109 (116–122), 2010. Santi, A., Amado, T.J.C., Acosta, J.A.A., 2003. Black oat biomass and nutrient cycling as affected by nitrogen fertilization in soil under notillage. Rev. Bras. Ciênc. Solo 27, 1075–1083. Silva, F.C., 2009. Manual of Chemical Analyzes of Soils, Plants and Fertilizers. Embrapa Informação Tecnológica, Brasília. BR, Brazil, pp. 191–233 (in Portuguese). Sultana, S.R., Ali, A., Ahmad, A., Mubeen, M., Zia-Ul-Haq, M., Ahmad, S., Ercisli, S., Jaafar, H.Z.E., 2014. Normalized difference vegetation index as a tool for wheat yield estimation: a case study from Faisalabad, Pakistan. Sci. World J. 14, 1–8. Tarnavsky, E., Garrigues, S., Brown, M.E., 2008. Multiscale geostatistical analysis of

621